➤ Modelling and analysis of the u Brumadinho tailings disaster using advanced geospatial analytics I. Atif1, F.T. Cawood1, and M.A. Mahboob2 Affiliation: 1 Wits Mining Institute (WMI), University of the Witwatersrand, Synopsis Johannesburg, South Africa. On 25 January 2019, one of the most significant and deadliest tailings dam failures in history occurred at 2 Sibanye-Stillwater Digital Brumadinho Córrego do Feijão iron ore mine in Brazil. Twelve million cubic metres of tailings travelling at Mining Laboratory (DigiMine), 120 km/h destroyed a total of 109 buildings, 36 belonging to Vale and 73 local residences. More than 259 Wits Mining Institute (WMI), people died. Some farmlands were wiped out and left under a sea of mud and tailings up to 8 m deep. Seven University of the Witwatersrand, sections of local roads, one main road, and one railway bridge were severely damaged. In this research, Johannesburg, South Africa. a GIS-based tailings spill path (TSP) model was developed using the Python programming language for predicting the potential tailings flow path – before failure of the tailings storage facility (TSF). The Correspondence to: pre- and post-failure satellite images of the Brumadinho disaster were processed and analysed to map I. Atif the damaged infrastructure and to extract digital footprints of the tailings waste and flow path. This model was then compared with the post-failure satellite images. The TSP model is capable of generating the possible path of tailings flow and other important outputs like a processed digital elevation model Email: (DEM), processed satellite image, down-path slope direction, and flow accumulation. The model was [email protected] tested and validated for the Brumadinho and Samarco tailing disasters. The results are very promising [email protected] and correlate well with the actual tailings spills. The methodology adopted in this research is robust, advanced, and can be applied to other tailings dams for hazard and risk assessment in case of their Dates: possible failure. The lack of high-resolution post-disaster satellite images and other topographical data Received: 26 Apr. 2020 were the main limitations of this research, which if available, could improve the modelling results. Revised: 27 Jul. 2020 Accepted: 3 Aug. 2020 Keywords Published: July 2020 Geospatial modelling, tailings dam failure, Brazil dam collapse, mining disaster, Brumadinho, Samarco tailings, tailings management. How to cite: Atif, I., Cawood, F.T., and Mahboob, M.A. Modelling and analysis of the Brumadinho tailings disaster using advanced geospatial analyticst. Introduction The Southern African Institute of The mining industry plays a significant role in supporting the economies of several developed and Mining and Metallurgy underdeveloped countries around the globe. Massive volumes of solid and liquid waste are produced DOI ID: as part of the mining and metals extraction processes. This has the potential for several negative http://dx.doi.org/10.17159/2411- socio-economic and environmental impacts. Among these, the greatest threat is failure of a tailings 9717/1196/2020 storage facility (TSF), which may contain a large volume of mining wastes. TSF failure is not a new phenomenon and many such events have occurred around the world, as shown in Figure 1. Some of the critical failures are Mount Polley in Canada (Byrne et al., 2015), Merriespruit in South Africa (van Niekerk and Viljoen, 2005), Cieneguita mine in Mexico (Warden, 2018), Cerro Negro in Chile (Valenzuela, 2018), Rio Pomba Cataguases in Brazil (Oliveira and Kerbany, 2016), and Bento Rodrigues in Brazil (Segura et al., 2016), but there are many more. There can be many reasons for TSF failures, such as an earthquakes (17% of global failures) (Villavicencio et al., 2014; Lyu et al., 2019), heavy rainfall (Ozkan and Ipekoglu, 2002), construction issues (17.3% of global failures) (Davies, 2002; Lyu et al., 2019), poor maintenance (Rico et al., 2008), excess pore water pressure (21.6% of global failures) (Wang et al., 2016, Lyu et al., 2019), starter wall and foundation failure and slope instability (Davies, 2002). Hence, proper monitoring, management, and risk assessment should be done to minimize the devastating impacts of a possible failure. To assess the risk of a potential TSF hazard, the first (and the most important) step is to map and quantify the magnitude of potential damage that can occur in the zone of influence. The zone of influence is the area that would be significantly affected in case of a TSF failure and should be categorized as a risk zone. Usually, field-based surveys can be used to map the zone of influence. However, such mapping is expensive and time-consuming. Recent advances in digital technologies like Geographical Information Systems (GIS) and Remote Sensing (RS) have made mapping and modelling a viable option for The Journal of the Southern African Institute of Mining and Metallurgy VOLUME 120 JULY 2020 405 ◀ Modelling and analysis of the Brumadinho tailings disaster using advanced geospatial analytics that 27 of the total number of TSFs in the country represent a high risk to human life, infrastructure, and the environment (New York TImes, 2019). If these TSFs, which are located in mountainous regions, failed, then an estimated number of more than 100 000 people living downstream could be affected, along with severe environmental destruction. Hence, for effective TSF management, it is essential to define the zone of influence as part of the risk assessment process. In this research, advanced geospatial analytics have been used to develop a model for mapping the potential tailings spill path (TSP) in case of a TSF failure. This can help decision-makers to assess, quantify, and manage the risk in a better, efficient, and scientific way. Figure 1—Historical global tailing dam failures. The colours correspond to Materials and methods the geographical region and numbers in the cells represent the number of accidents in that region for that decade (data source: Lyu et al., 2019) Study area The study area is a tailings dam at Brumadinho Córrego do Feijão iron ore mine, located in Belo Horizonte’s Brumadinho delineating the zone of influence. In the last decade, GIS and metropolitan district in southeastern Brazil (Figure 2). The RS have been extensively and successfully used in disaster average minimum and maximum temperatures of the study area management studies, e.g. seismic disasters (Ehrlich et al., are 16°C and 27°C respectively, with average annual precipitation 2009; Frigerio et al., 2016); flood disasters (Atif, Mahboob, and of 965 mm (38 inches). The Brumadinho region has a complex Waheed, 2015, Sajjad et al., 2020); geological disasters (Li et geology with different types of sedimentary ores like high- al., 2005, Mahboob et al., 2015b, 2019b); droughts (Raut et grade compact and soft haematite, soft itabirite, canga rica, al., 2020, Atif, Iqbal, and Su, 2019); and fire disasters (Hinkley, slumped canga, and rolado, with alluvial pebbles and cobbles 2019). Several commercial and open-source satellite data-sets of compact haematite (Simmons, 1968). Extensive mining has are available for both pre- and post-disaster analysis and future been conducted for more than 150 years (Chase, 2008) for gold risk assessments (Mahboob et al., 2019b; Voigt et al., 2007; van and iron. Brazil is the world’s second-largest iron ore producing Westen, 2013). country (De Moraes and Ribeiro, 2019). The applications of these advanced technologies can also be Upstream construction method useful in the mining industry, particularly for TSF management. The Brumadinho region has several tailings dams located in the For example, Rudorff et al. (2018) investigated the impact vicinity of mining areas, most of which are constructed using the of the Samarco tailings dam collapse on the turbidity of the upstream method (Valenzuela, 2018). Doce River plume off the eastern Brazilian coast by applying This economical construction method is very common in low Landsat and MODIS-Aqua imagery. Mura et al. (2018) used seismic risk areas. A cross-section illustrating the construction advanced differential interferometric synthetic aperture radar method is shown in Figure 3 (Soares, Arnez, and Hennies, (A-DInSAR) data from TerraSAR-X satellites to monitor the 2000). In this method, the fresh tailings are deposited on top spatial and temporal displacement and assess the vulnerability of the previously placed tailings. This requires the tailings to of the TSF at Samarco in Brazil. Goff et al. (2019) proposed provide support for the dam and as such makes this type of dam a cost-effective solution for monitoring and management potentially more prone to failure where the tailings are of low of tailings by combining satellite-based Earth observations strength and materials may be prone to liquefaction (McLeod and global navigation satellite systems, such as the Global Positioning System (GPS) technologies, with real-time on-site instrumentation. Wang et al. (2018) modelled tailings slurry runout using a Smoothed Particle Hydrodynamics (SPH) method. They concluded that SPH numerical modelling is a powerful technique that can be recommended for risk assessment and design assessments of TSFs. However, these numerical models are usually expensive and computationally very demanding when applied to solve the mathematical equations in order to predict the tailings flow paths. On 25 January 2019, one of the most severe tailing dam failures ever occurred in Brazil at Brumadinho